Automated Labour Detection Framework to Monitor Pregnant Women with a High Risk of Premature Labour

Abstract

It is estimated that more than 1 in 10 babies are born prematurely worldwide. Babies that survive premature birth are more likely to face lifelong health-related disabilities. By monitoring uterine contractions, labour can be detected, which assists in reducing premature birth complications.

Several studies have been conducted on monitoring pregnant women with a high risk of premature birth. The first study focused on home monitoring of uterine activity versus nurses' frequent contact with pregnant women. The second study compared \glsdesc{EHG} with \glsdesc{IUPC} to monitor pregnant women by recruiting 32 pregnant women in labour for a minimum of 30 minutes and used a simple algorithm to automatically recognize uterine contractions. The third study took randomized control trials of home uterine activity monitoring for pregnant women with a high risk of premature birth from 15 studies to determine if home monitoring systems can be used to evaluate pregnancy health status. The last three studies individually proposed the use of home mobile healthcare systems to monitor pregnant women.

Machine learning techniques have recently been used to predict and detect premature labour. Recent studies have used machine learning classifiers such as Random Forest and Decision Tree to categorize and recognize electrohysterography contractions with a high accuracy rate. In addition, deep learning models such as artificial neural networks, similar to machine learning techniques, have been designed to mimic the human brain to analyze and extract complex relationships between data.

In this research, we aim to mitigate the consequences of premature birth for pregnant women and the fetus by proposing a safe, simple, home-comfortable, low-cost, and reliable monitoring framework. The system uses a non-invasive method to monitor uterine electrohysterography contractions using a wireless body sensor and a smartphone. The smartphone will analyze uterine readings, and if they contain a premature labour pattern, a warning notification will be triggered. The framework will have three schemes: an amplitude-frequency algorithm scheme, a machine learning algorithm scheme, and a deep learning scheme. A proof-of-concept prototype was designed and tested for reliability, performance and power consumption using five electrohysterography uterine contraction databases. The results show that the schemes were able to meet the framework’s objectives.